Está en la página 1de 3

Atmospheric Environment 36 (2002) 2263–2265

Letter to the editor

Comments on: Using CALPUFF to evaluate the impacts
of power plant emissions in Illinois: model sensitivity
and implications
Michael R. Amesa, Stephen G. Zembaa, Robert J. Yamartinob, Peter A. Valbergc,
Laura C. Greena
a

Cambridge Environmental Inc., 58 Charles St., Cambridge, MA 02141, USA
b
Integrals Unlimited, 509 Chandler’s Wharf, Portland, ME 04101, USA
c
Gradient Corporation, 238 Main Street, Cambridge, MA 02142, USA

Levy et al. (2002) describe portions of their analysis of
the public health impacts of emissions from nine power
plants in Illinois. This and similar analyses (Levy and
Spengler, 2002; Freeman, 2001; Abt, 2000) use dataintensive models of air quality and health risks to
estimate the number of deaths caused by increases in
ambient particulate matter (PM) concentrations due to
emissions from specific sources. As modeled by Levy
et al. (2002), 87% of the population-weighted, powerplant-derived PM2.5 increment exists as sub-microgram
per cubic meter concentrations of secondary sulfates and
nitrates, which, together with modeled primary PM, are
estimated to cause ‘‘320 premature deaths per year y
due to current emissions from the nine Illinois power
plants’’. As explained below, several crucial but difficultto-recognize problems with the Levy et al. (2002)
analysis compromise its reliability. We thank Dr. Levy
and coworkers for sharing with us the input files that
control the CALMET and CALPUFF runs, that we
might explore these issues.

1. Atmospheric dispersion and transport modeling issues
Chicago is the largest population center in the
modeled region, home to two of the nine power plants,
and generally downwind of six of the other seven
facilities modeled. The city’s location on the shore of
Lake Michigan leads to sharp changes in meteorological
conditions due to abrupt lake/land and rural/urban
effects. The lake also reflects a step-function in the
population distribution that could magnify the effects of
any modeling errors in this region. Thus, the precise
E-mail address: ames@cambridgeenvironmental.com
(M.R. Ames).

pattern of the near-field dispersion estimates can
profoundly influence the study’s results.
Levy and co-workers’ initial report (2000) lists
distances from each plant at which the maximum
ground level concentrations (MGLC) of primary PM
are predicted. These distances range from 1.1 to 40 km.
As expected, plants with taller plume heights (stack
height plus modeled plume rise) yielded greater estimated distances to the MGLC. However, the two
modeled plants within Chicago (Fisk and Crawford)
were found to have much shorter distances to primary
PM MGLC (1.1 and 1.6 km, respectively), and consequently much higher GLC values than the other sources,
despite similarities in plume heights. Although near-field
emissions from these two plants are modeled using fastgrowing urban dispersion coefficients in contrast to the
slower-growing rural coefficients (Scire et al., 2000) used
for the other plants, this difference alone cannot account
for the differences in estimates of distance to MGLC.
However, choices made in generating the driving
CALMET meteorological data could strongly influence
the results of these CALPUFF dispersion calculations.
CALMET uses several different algorithms (Venkatram, 1980; Maul, 1980; Zilitinkevich, 1972) to predict
mixing heights that utilize micro-meteorological variables in a given cell, but do not recognize effects of
horizontal advection. To avoid unrealistic jumps between adjacent cells and to attempt to account for
advection, CALMET spatially averages mixing heights
based on an upwind-looking window, coupled with a
user-specified, isotropic, averaging window of radius
MNMDAV cells. The MNMDAV default value is 1, but
Levy et al. chose to override this with MNMDAV=3—
a value that translates to a 45 km (Levy, 2001) radius
smoothing window that is large compared with mechanisms responsible for spatial smoothing of mixing depths.

1352-2310/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 2 - 2 3 1 0 ( 0 2 ) 0 0 2 4 6 - 7

Given that the effect of this phenomenon is not as evident on the Waukegan plant. inclusion of near-lake surface station observations can force the near-surface winds to respond to these effects. However. this averaging near the Western shore of Lake Michigan dramatically smooths the otherwise rapid transition from relatively high. 2. Since the Fisk and Waukegan plants are located in a modeling cell adjacent to a water cell.. for a buoyant plume near its source. The CALMET meteorological model was run by Levy et al.2264 M. They also write that the concentration-response function for excess mortality that they apply (Krewski et al. 3. over-water mixing depths. 2001). As the authors themselves note: ‘‘yit appears likely that the degree of uncertainty in atmospheric modeling will not dominate the total uncertainty associated with health impact or benefit estimation’’. While this resolution is generally appropriate for modeling large. 1998). Health effects modeling issues The uncertainties associated with the modeling of primary and secondary PM from power plant emissions are substantial. it may be that the combination of this mixing height reduction and the use of urban dispersion curves conspire to erroneously predict that the Fisk and Crawford plant plumes reach the ground at unreasonably short distances. but the basic meteorological model employed to drive CALMET was developed on a grid with a 40-km lateral resolution.. and no one knows which of these substances. hasten mortality (Moolgavkar and Luebeck. Although the choice of a 15-km grid resolution for CALMET significantly limits the detail with which one can capture lake/land breezes. chose to include only one surface meteorological observing station (Levy. Neither was utilized by Levy et al. Even though the 15 km resolution of the CALMET fields precluded explicit representation of lake/land breezes. Ambient PM is a mixture of thousands of substances that vary in physical. 1998. near the coastline of Lake Michigan. Levy et al. experimental evidence on these compounds led the Netherlands Aerosol Program (NAP. 2002). let alone fatal. 2001). which casts further uncertainty on the near-field predictions made for sources located near the lake/land interface. We do. few of the enhanced capabilities being added to models have received thorough evaluation on even an isolated basis. These are understatements. the toxicologic. 1996. However. The authors indicate that the NOAA RUC2 40-km resolution data were augmented by the ARPS Data Assimilation System (ADAS) and ACARS aircraft-reported wind and temperature data. Ames et al. moving beyond the simulation of unidirectional convection and dispersion of passive tracer species to simulating many atmospheric processes. a lower inversion height brings the plume down much more quickly.R. / Atmospheric Environment 36 (2002) 2263–2265 Importantly. As discussed elsewhere (Green et al. the 15-km grid and especially the 40-km grid are too coarse to resolve important flows. relatively flat domains. 1998. however. have a good idea as to what fractions of ambient PM are not likely to be toxic. overland mixing depths to shallow. the universal applicability of the simplistic correlation-based . there are uncertainties in each of its many components (e. emissions from these facilities are subject to dramatic reductions in mixing height due to this spatial averaging. leading to a shorter distance to MGLC and. and causes mixing heights over Chicago to be reduced significantly. and biological properties. Conclusions Atmospheric modeling is an important tool to assess the potential impacts of various sources of pollution. let alone as part of an integrated modeling system. Valberg and Watson. to many-fold higher values of GLC in the highly populated Chicago urban area. to conclude that ammonium sulfate and ammonium nitrate ‘‘seem[s] to be toxicologically inert at current concentrations y Decreasing the levels of inert compounds will not reduce the health risk of the population’’. chemical. these reductions lead only to reciprocal increases in MGLC that are generally modest. We recommend examination of the relative importance of these various factors through scrutiny of the extensive CALMET meteorological data files and/or appropriate CALMET/CALPUFF sensitivity runs. CALPUFF modeling options could have been used to: (1) introduce a detailed coast line to better quantify source-to-coast line distances and (2) activate a module to simulate plume dispersion within (and above) a computed thermal internal boundary layer. or that the Waukegan emissions are often injected into the stable layer above these low mixing heights. Gamble. such as lake/land breezes. with a spatial resolution of 15 km. 2000) is ‘‘quite uncertain and has numerous issues associated with its implementation’’. 2001) out of the dozens available within the modeling domain. and that station is about six grid cells (or 90 km) inland from the Western shore of Lake Michigan—a distance far too great to record lake/land breezes. and these fractions include secondary sulfates and nitrates. if any. in this case. (Levy. For wellmixed plumes.g. for example.. Modeling capabilities have expanded greatly in recent years. Although the CALMET/CALPUFF system is generally based upon first principles. and the Crawford plant is only two cells from the water. Phalen. yet they pale when compared to the central uncertainties of the health-risk model used by Levy et al. so perhaps the different types and quality of data play important roles over the Waukegan versus Chicago areas.

J.. 1773–1782. Golomb. C. Kumar.. 5–18.. F. J.I. 2000.. S. interagency workgroup on air quality modeling (IWAQM) phase 2 summary report and recommendations for modeling long range transport impacts. R. Personal Communication of Model Input Data. if anything.. MA.. Bowdoin College. R. Moolgavkar.T. Bilthoven. Siemiatycki. Sullivan. Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of Particulate Air Pollution and Mortality. Offsetting errors among various processes and source terms may mask large uncertainties associated with the modeling of individual sources and species.. Cambridge.M. 1999. Moon. J. M. pollution-derived PM in the United States accelerates death. Nottingham. 2002) end their paper by writing.F.. 2001. C. Health Effects Institute. Financial support was provided by a grant from MidWest Generation. E. Since the near-field predictions are a critical determinant of populationweighted predictions in the urban Chicago setting. Discussion Document on Health Risks of Particulate Matter in Ambient Air. 2002. MA. ambient. .. Fay et al. Central Electricity Generating Bureau MID/SSD/80/0026/R. P. As the bulk of CALMET/CALPUFF’s validation has focused on long-range transport of passive tracers (US EPA. Toxicology Letters 96–97. Concord. A. R. MA. Brunswick. Levy... White. Freeman III.. ME. A. 1980. S.. 1999). L.. 1998. / Atmospheric Environment 36 (2002) 2263–2265 MESOPUFF chemical mechanism). one cannot gauge the reliability of the model to predict incremental secondary PM concentrations from specific power plant emissions of SO2 and NOx. Using CALPUFF to evaluate the impacts of power plant emissions in Illinois: model sensitivity and implications.D. EPA-454-R-98019. Inhalation Toxicology 10. study.. 1972. 1985. MA. 1996. Uncertainties relating to the health effects of particulate air pollution: the US EPA’s particle standard. D. Spengler... Atmospheric Transport of Sulfur Compound Pollutants. Alternative hypotheses for PM associations with daily mortality and morbidity.. Department of Environmental Health.. S. A User’s Guide for the CALMET Meteorological Model (Version 5). D. Modeling the benefit of power plant emission controls in Massachusetts.. M. J.E. Luebeck. 2001. CA. Atmospheric Environment 36. March 19–22...S.A. Hlinka.. Source apportionment of wet sulfate deposition in eastern North America.. Final Draft. A. 1998. Valberg. (1985) used a simple model to produce reasonable estimates of acid deposition.Y. D. 1063–1075. Burnett. M. 2001. Earth Tech.. Phalen.5 and mortality in long-term prospective cohort studies: cause-effect or statistical associations? Environmental Health Perspectives 106.H. PM2. Watson. L. Spengler. Department of Environmental Health. Krewski.G. The authors (Levy et al.. MA.. Robe. the uncertainty of the specific numerical estimates is much greater than indicated by the authors. Journal of the Air and Waste Management Association 52 (1).. Abrahamowicz. Gamble. but could construct similar results by counter-varying the importance of processes such as SO2 to sulfate conversion and dry/wet deposition. 641–662. 263–267. Crouch. we question whether these multi-layered modeling exercises form a reliable basis for public health policy making. M. J. A.R. Spengler. F.. RIVM. 2000. Fay.. Boundary Layer Meteorology 3. M. Los Angeles. K. and especially its ability to estimate detailed spatial patterns of individual species. 141–145. Acknowledgements We thank Jonathan Levy and his colleagues for sharing their underlying data and analyses. D. Ames. D. Maul. R. England. The Particulate-Related Health Benefits of Reducing Power Plant Emissions.. Boston. This inherent modeling uncertainty is exacerbated by the specific issues we have identified with near-field predictions in the Levy et al. Zilitinkevich.. J.. in press. Ames et al. Atmospheric Environment 19 (11).H. D. Scire. An Analysis of the Pataki Power Plant Proposal vs. Hlinka. 1980. Hoover.. References Abt Associates Inc. R.S. 2002. In: Eleventh NATO-CCMS International Technical Meeting on Air Pollution Modeling and its Application. Levy. Goldberg. Inc. W. 2002.. Harvard School of Public Health. Clean Air Task Force. What’s wrong with the national ambient air quality standard (NAAQS) for fine particulate matter (PM2. Boston... United States Environmental Protection Agency (US EPA).. Jerrett.. M. Netherlands Aerosol Program (NAP).R. Epidemiology 7 (4). J. Estimated Public Health Impacts of Criteria Pollutant Air Emissions from Nine Fossil-Fueled Power Plants in Illinois. 535–549.. J. Yamartino.. P.. in current. J. Full Environmental Comparability Using 2265 the New York Externalities Cost Model. 420–428.. Venkatram. Fernau. Green. D. A critical review of the evidence on particulate air pollution and mortality. Levy. Levy.I. 2000. Sullivan. On the determination of the height of the Ekman boundary layer. Given both (i) persistent uncertainties in the air modeling and (ii) how little we understand about what.5)? Regulatory Toxicology and Pharmacology.. ‘‘The magnitude of the public health impacts associated with these concentration increments is potentially significant and illustrates that accurate long-range dispersion modeling can provide meaningful and policy-relevant information for the regulatory community’’. Much of the analysis presented above could not have been performed without their collegiality. J. D. Estimation of turbulence velocity scales in the stable and unstable boundary layer for dispersion applications. T. D.S. Lash.. The Netherlands. Harvard School of Public Health. Boston. J. et al. E. 1998. 2000.